Quoting Joonas Lahtinen (2020-11-16 10:22:23) > Quoting Huang, Sean Z (2020-11-15 23:07:53) > > Enable one ioctl action to allow ring3 driver to set its ring3 > > context, so ring0 PXP can track the context id through this ring3 > > context list. > > Overall the patches should refer to "userspace" not "ring3" to avoid > confusion. "kernel" vs "user" not ring0 vs ring3. There's also a missing chunk as to why this is not associated with the existing user context, rather than introducing a new incomplete encapsulation. Overall, you've left in an awful lot of debug code and failed to follow the coding style. A confusion as to whether your hw interactions is on the GT or the whole device (it's GT). Wrt the flow of the patches, robust setup and termination must be early in the series, not tacked onto the end. And you seem to confuse the kernel contexts as something special, you use them as non-privileged, just like ordinary userspace. Do not use the engine->kernel_context! You risk breaking (and judging from the waits you do add, it is inevitable that you have broken) power management and heartbeats. -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
